> ## Documentation Index
> Fetch the complete documentation index at: https://docs.incredibuild.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Scaling Helpers in GCP

Last updated on Dec 22, 2025

## Overview

This topic explains the mechanics involved in scaling cloud Helpers automatically using Managed Instance Groups, for Incredibuild for Windows. The process involves these steps:

1. Create an Instance Template (one or more).

2. Create a Managed Instance Group (one or more).

Because of the way autoscaling works, you will need to have a single Helper machine alive in the VPC at all times. The Managed Instance Group spins up machines based on the load on the entire group, so one machine needs to be available as a bootstrap.

## Requirements

To use autoscaling you must have:

* A GCP account, with enough permissions.

* Access between your Coordinator and Initiators to the VPC where the helpers will run, over all relevant TCP ports (see [System Requirements](/windows/system-requirements))

It is assumed that:

* You have a working Incredibuild environment, either on-prem or in your VPC.

* You have a network connection between your Coordinator and Initiators to/from your VPC.

* You are experienced with Incredibuild technology.

* You are well versed in managing GCP VM instances.

## Step 1: Create an Instance Template

Create a Launch Template, using the following sample GCP `gcloud` command.

The following scripts assume that `COORD` contains the persistent IP address or FQDN of the Coordinator.

In the following script, replace the value of `COORD` with the persistent IP address or FQDN of the Coordinator, `INSTANCE` with the name of the machine type, and `VOL` with the size of the Helper disk (in GB).

[Copy](javascript:void\(0\);)

```
VOL="50" 
INSTANCE="n2-standard-4" 
COORD="192.168.1.1" 

gcloud beta compute instance-templates create helper-template \ 
--machine-type=$INSTANCE \ 
--network-interface=network=default,network-tier=PREMIUM,stack-type=IPV4_ONLY \ 
--create-disk=auto-delete=yes,boot=yes,device-name=helper-template,image=projects/windows-cloud/global/images/windows-server-2022-dc-v20241212,mode=rw,size=$VOL,type=pd-balanced \ 
--provisioning-model=SPOT \ 
--metadata=windows-startup-script-ps1="\$COORD="$COORD";   \$ProgressPreference = 'SilentlyContinue'; invoke-webrequest https://dl.incredibuild.com/ib10-latest-silent -outfile  C:\ibsetup.exe; invoke-command -ScriptBlock { C:\ibsetup.exe /install /COMPONENTS=Agent /AGENT:AGENTROLE=helper /AGENT:HELPERTYPE=floating /Agent:InstallAddins=off /COORDINATOR=$COORD};"
```

Note that the Launch Template needs to connect to the same Security Groups and Subnet as the original Helper.

## Step 2: Create a Managed Instance Group

The Managed Instance Group you create will control the range of vCPUS in the group (including thresholds when to spin up/down).

Create a Managed Instance Group, using the following sample GCP `gcloud` command.

In the following script, change the various Managed Instance Group options to match your grid acceleration requirements.

[Copy](javascript:void\(0\);)

```
gcloud beta compute instance-groups managed create helper-instance-group \ 
--base-instance-name=helper-instance-group \ 
--template=helper-template \ 
--size=1
```

Set the scaling policy for the Managed Instance Group, using the following sample GCP `gcloud` command.

[Copy](javascript:void\(0\);)

```
gcloud beta compute instance-groups managed set-autoscaling helper-instance-group \ 
--mode=on \ 
--min-num-replicas=1 \ 
--max-num-replicas=4 \ 
--target-cpu-utilization=0.5 \ 
--cpu-utilization-predictive-method=none \ 
--cool-down-period=60
```